Does anyone in the OMPI community want to work on a 6-function MPI
proof-of-concept for an MPI ABI on Linux, collaborating with Intel MPI?
At the MPI Forum meeting in Dublin, the MPI ABI meeting was... er...
shall we say, "spirited." :-) Both the benefits and drawbacks of an
MPI ABI are widely contended (it's a surprisingly complex topic).
Several items were decided by the end of the ABI meeting. The
following is an abbreviated list; see the ABI working group's summary
slide presentation for a more complete list:
--> In particular, see slide 8
- If it is ever completed, MPI ABI compliance will be a separate
entity from the MPI 2.x and 3.x standards. ABI compliance will be a
checkmark for an MPI implementation, but will be unrelated to an
implementation's 2.1, 2.2, 3.0, ...etc. compliance.
- Microsoft and Intel MPI will go off and try to implement a 6-
function MPI ABI morph layer for Windows-based MPI implementations.
The goal is to report back on the progress for the December MPI Forum
meeting (i.e., after SC08).
- Intel MPI has volunteered to work with another Linux MPI
implementation to do the same (proof of concept 6-function MPI ABI
morph layer for the Dec Forum meeting).
*** I was asked if Open MPI would be the 2nd Linux MPI (the Argonne/
MPICH crew has already said "no"). I said that I did not have any
cycles/priority to work on this, but that I would ask the rest of the
OMPI community to see if anyone else wanted to. Does anyone want to
work with Intel MPI on this project?
(we can discuss this on the call tomorrow in detail, if desired)